Sufficiency causatives Prerna Nadathur Department of Linguistics - - PowerPoint PPT Presentation

Sufficiency causatives

Prerna Nadathur

Department of Linguistics Stanford University March 14, 2019

joint work with Sven Lauer, University of Konstanz

The basic puzzle

Languages use a range of non-interchangeable periphrastic causatives:

(1)

• a. Gurung caused the children to dance.
• b. Gurung made the children dance.
• c. Gurung had the children dance.
• d. Gurung got the children to dance.

◮ each example describes a causal situation/chain of events:

◮ some action by or involving Gurung causally brought about an

event in which the children danced

◮ but (1a)-(1d) don’t describe the same situations

◮ (1b): force/coercion ◮ (1c): causer authority (no resistance) ◮ (1d): manipulation or bribery ◮ (1a): indirectness

The basic puzzle

Languages use a range of non-interchangeable periphrastic causatives:

(1)

• a. Gurung caused the children to dance.

[indirectness]

• b. Gurung made the children dance.

[coercion]

• c. Gurung had the children dance.

[authority]

• d. Gurung got the children to dance.

[manipulation]

Two questions:

• 1. What’s shared (semantically) between periphrastic

causatives? What produces the common causal meaning?

• 2. What’s different? What makes them sensitive to different

features of causal scenarios (volition, resistance, authority, etc)? Today: we’ll focus on make, comparing it to cause

A first hypothesis

Causative verbs share a common causal core of meaning, call it cause (Dowty 1979)

◮ cause ≈ cause ◮ cause is a basic semantic atom

◮ might be definable (e.g. in terms of counterfactuals or

necessity; Lewis 1973)

◮ doesn’t break down into further cause-related components

◮ to this core, different periphrastic verbs add different

non-causal entailments make = cause + coercive implication (2) X make Y do Z := X cause Y to Z + Y did not want to do Z

A first hypothesis

X make Y do Z := X cause Y to Z + Y did not want to do Z Problems:

◮ make is fine when the causee plausibly wants the outcome:

(3) “Then a surprise surgery and hospital stay at the age of 13 brought Albert in contact with nurses who made her feel happy and important during a stressful situation.”

◮ so, revise the coercive implication?

. . . if Y had not wanted Z(Y ) to occur, it still would have

◮ but: make is also felicitous with non-volitional causees:

(4) “Too much water made the plant die...” (5) “Mussolini made the trains run on time.”

A first hypothesis

On the cause-as-core approach for make:

◮ we want to derive the coercive implication for

make-causatives

◮ but we can’t make reference in the semantic representation to

the volitional state of the causee One way out: causative make is polysemous (Wierzbicka 1998)

◮ the coercive implication is specified for interpersonal make

(6) “[Anand’s mother] made Anand pump the tires [of the bicycle] every morning.” [Naipaul, A House for Mr. Biswas]

◮ but not for the impersonal make of surprise:

(7) The wind made the door slam shut.

◮ or the make of subjective necessity:

(8) “A sharp hiss made Alice draw back in a hurry.” [Carroll, Alice’s Adventures in Wonderland]

An alternative route?

The many makes route isn’t very satisfying:

◮ intuitively, make-sentences have something in common ◮ replacing make with other causatives produces changes in

meaning and felicity conditions:

(9) . . . brought Albert in contact with nurses who made her feel happy . . . = . . . brought Albert in contact with nurses who got her to feel happy . . . (10) Mussolini made the trains run on time. = Mussolini had the trains run on time.

Today:

◮ set aside the cause-as-core hypothesis ◮ central claim: causatives share causal meaning, but express

different types of causal dependencies

◮ specific implications (e.g. coercion) follow from the type of

dependency asserted

Sufficiency causatives

Sufficiency thesis: make is a sufficiency causative, expressing that a causing event made its effect inevitable

◮ make is neither a hyponym nor a hypernym of cause, but

expresses a different type of dependence

◮ causal dependence relations can be defined in a unified way, as

configurations in a causal network (Pearl 2000, Schulz 2011)

◮ the consequences of a sufficiency analysis for make: the

coercive implication

Cause, make, and counterfactual necessity

Cause plausibly predicates a counterfactual relationship between cause and effect (Lewis 1973):

(11) “In total, the fires caused the transit system to lose $68,000. . . ” → If the Napa fires had not occurred, the transit system would not have lost $68,000

While a counterfactual is often pragmatically plausible for make:

(7) The wind made the door slam shut. The door would not have slammed were it not for the wind.

. . . there are felicitous uses which explicitly deny necessity:

(12) I usually go to soccer camp in the summer. Last year I was thinking about going to band camp instead, and I could not make up my

• mind. Then I broke my ankle, which settled things. I am so happy

the injury made me skip soccer camp. I had the best summer ever! I would have gone to soccer if I hadn’t broken my ankle.

Cause, make, and counterfactual necessity

Make is not a hyponym of cause:

◮ cause is bad in the soccer-camp scenario

(13) I usually go to soccer camp in the summer. Last year I was thinking about going to band camp instead, and I could not make up my mind. Then I broke my ankle, which settled

• things. ??I am so happy the injury caused me to skip soccer
• camp. I had the best summer ever!

◮ this suggests: cause is associated with (counterfactual)

necessity, while make is not Pursuing the sufficiency thesis:

◮ make is good in the soccer camp scenario because the injury

“settles” things.

Make, cause, and causal sufficiency

We can test this intuition with a slightly different scenario:

(14) Several things happened last summer which led me to skip tennis

• camp. First, I broke my ankle in the spring, and since it was taking

a long time to heal, I started thinking about band camp for the first

• time. Then I got into an argument with my doubles partner, so even

with my ankle getting better, I wasn’t sure I wanted to go to tennis. Finally, my parents said they’d get me a trombone if I went to band camp, which was pretty tempting!

• a. ?I am so happy the injury made me go to band camp! I had

the best summer ever.

• b. I am so happy the injury caused me to go to band camp! I

had the best summer ever.

• c. → Breaking my ankle made it inevitable that I would go to

band camp. [sufficiency not supported]

Causal sufficiency: positive consequences

Sufficiency thesis: make asserts that the indicated cause was causally suf- ficient for the effect: given the cause, the effect was guaranteed. X make Y do Z := X ensured that Y Zed We do not directly encode the coercive implication:

◮ but, the sufficiency analysis should produce it naturally when

the embedded VP is a volitional action

(1b) Gurung made the children dance. → Gurung’s action guaranteed that the children danced.

◮ if the children acted freely in dancing, then Gurung’s action

couldn’t have made the dancing inevitable (they could have changed their minds and not danced)

A bonus: necessity, pragmatically

Sufficiency thesis: X make Y do Z := X ensured that Y Zed Question: why is the idea that make predicates necessity so prevalent?

◮ there is a well-known tendency for sufficiency statements to

be interpreted as conveying necessity

◮ conditional perfection: if P, then Q is often interpreted as if

and only if P, then Q (Geis & Zwicky 1971)

(15) If you study for the exam, you’ll get an A. If you don’t study, you won’t get an A

◮ but conditional perfection is defeasible (cancellable):

(16) If you study for the exam, you’ll get an A. Actually, you might get an A even if you don’t study.

A bonus: causal perfection

Claim: if make predicates sufficiency, then necessity implications can arise pragmatically as causal perfection implicatures

◮ if so, we expect:

(a) contexts where necessity arises as part of the speaker’s intended meaning: ‘exculpatory’ uses of make

(17) The devil made me do it. → I had no choice but to do it [coercion/sufficiency] I would not have done it else [perfection/necessity]

(b) contexts that cancel necessity inferences

(18) My husband’s arrest (finally) made me get a divorce. . . . Even if his arrest had not made me do it, I might have gotten a divorce anyway, given the way he treated me.

Interim summary

Causative verbs differ in the type of causal dependence they assert:

◮ make is a sufficiency causative ◮ . . . allows us to capture the coercive implication without

hard-coding it

◮ . . . gives us a handle on apparent necessity inferences ◮ . . . explains why make and cause contrast ◮ because cause asserts (causal) necessity (and maybe some

• ther stuff, but crucially not sufficiency)

Causal dynamics and causal dependency relations

We cash out causal dependencies in a causal model (Schulz 2011, Pearl 2000)

◮ a dynamics is a graphical model:

◮ nodes are events or propositions, and can take on truth values

(0, 1, undetermined)

◮ arrows represent causal relevance links (P → Q if P is a causal

influencer of Q)

◮ it comes along with a set of equations defining the causal

links

◮ given an initial setting for the nodes, we can use these

equations to calculate causal consequences (normal causal developments)

◮ main idea: causal dependence relations (necessity,

sufficiency) are labels for certain structural configurations in a dynamics

◮ these labels appear as atoms in the semantics of causative

verbs

The Lifschitz example (1990)

(19) The circuit example:

• a. Suppose there is a circuit with two switches and one light,

such that the light is on (L) exactly when both switches are in the same position (up or not up).

• b. At the moment switch 1 is down, and switch 2 is up.

◮ (a) states the causal laws (dynamics) ◮ (b) gives us an initial setting (background situation) ◮ given (b), a normal causal development will be a situation

in which the light is off (L = 0)

Causal necessity and causal sufficiency

Given two events C and E, and a background situation s which does not fix the occurrence of C . . . (20) C is causally sufficient for E relative to s if

• a. s does not produce E as a normal causal development

the effect wasn’t already inevitable

• b. s′ = s + C does produce E as a normal causal development

the cause guarantees the effect

(21) C is causally necessary for E relative to s if

• a. s does not guarantee E
• b. s′ = s + C has a supersituation s′′ which does not fix E, but

has it as a normal causal development the cause makes the effect possible

• c. there is no supersituation s′′ of s′ which makes (b) true but

does not have C as a normal causal development the effect was not possible without the cause

The Lifschitz example again

Suppose switch 1 is fixed up (S1 = 1). In this background situation, flipping switch 2 up is both necessary and sufficient for the light to come on.

◮ if make predicates sufficiency and cause predicates necessity,

we correctly predict that . . . (22)

• a. Turning the second switch on made the light go on.
• b. Turning the second switch on caused the light to go on.

. . . are both acceptable

A necessary but insufficient condition

The fire scenario

A fire burned a field outside town, starting immediately after the electric company restored power to a line that ran through the field, and which had previously been shut off. Due to a long drought, the grass was unusually dry, and thus more susceptible to burning. We assume that a power line can spark a fire only if electricity is flowing and the line comes into contact with something grounded and inflammable. Unfortunately, there had not been an inspection of the area for several months, and so the condition of the line is unknown, and cannot be determined by direct evidence in the aftermath of the fire. P= power restored in line D = drought conditions G = dry/inflammable grass L = downed line F = fire

A necessary but insufficient condition: the fire scenario

(23)

• a. #Restoring power made the field catch fire.
• b. Restoring power caused the field to catch fire.

We capture these judgements:

◮ background situation: drought, dry grass (D = G = 1) ◮ P is a necessary but insufficient condition for F

◮ necessity: there’s no route from the background to the fire

without P

◮ sufficiency: we don’t guarantee F until we know L

P= power restored in line D = drought conditions G = dry/inflammable grass L = downed line F = fire

A sufficient but unnecessary condition

The bus scenario

Lia has a high-end triathlon bike, which she rides to work almost every

• day. She takes the bus when rain is predicted for the evening, because it

gets too cold to bike home in the rain. Occasionally, Lia’s friend Ava visits her. Ava is a pro cyclist. She gets up early and borrows Lia’s bike for training when she has a race coming up. At the moment, Ava is visiting. She has a race coming up in two weeks, and is in the middle of training. Rain is predicted for the evening. V = Ava visiting T = Ava training G = bike is gone R = rain is forecast B = Lia takes the bus

A sufficient but unnecessary condition: the bus scenario

(24)

• a. Ava’s training made Lia take the bus to work.
• b. #Ava’s training caused Lia to take the bus to work.

◮ background situation: Ava is visiting, rain is forecast

(V = R = 1)

◮ T is sufficient but unnecessary for B:

◮ since rain is forecast, we can get B = 1 with G = 0 and

without T = 1 (unnecessary)

◮ T guarantees B

V = Ava visiting T = Ava training G = bike is gone R = rain is forecast B = Lia takes the bus

Refining the account: temporal considerations

The lighthouse scenario

The lighthouse was built with a very sturdy foundation, designed to withstand high winds at the tower top, but the foundation sustained structural damage in an earthquake about ten years ago. Even that would have been fine, but this year, there were record-setting winds and the worst hurricane season anyone can remember, and given the prior damage, it could not take the extra strain. Q= earthquake S = storms L = tower collapses

Refining the account: temporal considerations

Q= earthquake S = storms L = tower collapses

(25)

• a. The earthquake caused the tower to collapse.
• b. The storms caused the tower to collapse.
• c. #The earthquake made the tower collapse.
• d. The storms made the tower collapse.

. . . we can’t rule out (25c) right now!

◮ intuitively: the earthquake isn’t sufficient because things might still

have gone another way

(26) Temporal constraint on background situations. In evaluating

causal claims, the background situation cannot contain facts that were not fixed at the time of the purported cause.

Other constraints: revisiting the coercive implication

Sufficiency intuitively captures the coercive implication: X make Y do Z Y did not make a free decision to do Z

◮ but, consider the following:

(27) The children are eager to dance. Gurung is their (strict) instructor, and they are only allowed to dance if he gives his explicit permission. Finally he does so, and they dance happily.

• a. ??Gurung made the children dance.
• b. Gurung let the children dance.

◮ the coercive implication doesn’t follow, because nothing stops

the children from changing their minds

Sufficiency and volition: command scenarios

(28) The children are tired, and are not looking forward to dancing in their dance class. Gurung is their strict instructor, and they are required to obey any commands he gives them. As soon as they arrive in class, he tells them to dance. They start dancing.

• a. Gurung made the children dance.
• b. ??Gurung let the children dance.

The judgements are the same even if the children want to dance:

(29) The children have been feeling eager to dance all afternoon, looking forward to their lesson. Gurung is their strict instructor, and they are required to obey any commands he gives them. As soon as they arrive, he tells them to dance. They immediately start dancing.

Sufficiency and volition: presupposition

Make is sensitive to the relationship between the volition of an agent (when there is one), and the stated cause:

◮ it picks out contexts where the will of the agent is irrelevant

◮ either because the agent does not have control (manipulation,

trickery, persuasion)

◮ or where the cause renders the agent’s will irrelevant

(command)

◮ but rules out contexts where a cause is sufficient only in view

• f the agent’s own goals (permission)

(30) Constraint on volitional actions. In evaluating a make-causative with background situation s, causing event C, and caused event E, a proposition WE representing the agent’s intention to perform E can either be determinative of E relative to s + C or fixed by s/C, but not both.

Sufficiency and volition

Constraint on volitional actions. In evaluating a make-causative with background situation s, causing event C, and volitional caused event E, WE can either be determina- tive of E relative to s + C or fixed by s/C, but not both.

X

• ◮ this captures coercion

◮ hypothesis: let is a sufficiency

causatives which picks out complementary configurations

◮ intuitively, force, enable,

permit are sensitive to similar constraints

Summary

◮ we can maintain a common notion of causation without

cause as the only causal semantics atom

◮ claim: causatives predicate different dependencies ◮ preliminary account of make as a sufficiency causative

◮ a formal way to cash this out ◮ positive consequences for specific implications

◮ looking ahead: finer-grained distinctions between causatives

• n the basis of what kinds of background constraints

(presuppositions)

◮ e.g., permissive causations (enable, let, permit) care about

configuration of volition and cause

◮ coercive causatives (make, force) are complementary ◮ languages will differ: German lassen is a ‘pure’ sufficiency

causative, with both let and make meanings (Lauer & Nadathur 2018)

◮ causal models give us a way of capturing these distinctions

while developing unified accounts of periphrastic causatives

References

• 1. Dowty, D. 1979. Word meaning and Montague Grammar. Dordrecht:

Reidel.

• 2. Geis, M. & A. Zwicky. 1971. On invited inferences. Linguistic Inquiry 2.
• 3. Hobbs, J. 2005. Toward a useful concept of causality for lexical
• semantics. Journal of Semantics 22.
• 4. Lauer, S. & P. Nadathur. 2019, under review. Causal necessity, causal

sufficiency, and the implications of causative verbs.

• 5. Lauer, S. & P. Nadathur. 2018, in revision. Sufficiency causatives.
• 6. Lewis, D. 1973. Causation. Journal of Philosophy 70.
• 7. Lifschitz, V. 1990. Frames in the space of situations. Artificial

Intelligence 46.

• 8. Pearl, J. 2000. Causality. Cambridge: Cambridge University Press.
• 9. Schulz, K. 2011. If you’d wiggled A, then B would’ve changed: causality

and counterfactual conditionals. Synthese 179.

• 10. Shibatani, M. 1976. The grammar of causative constructions: a
• conspectus. In The Grammar of Causative Constructions, Shibatani (ed).

New York: Academic Press.

• 11. Wierzbicka, A. 1998. The semantics of English causative constructions in

a universal-typological perspective. In The New Psychology of Language, Tomasello (ed). Mahwah, NJ: Lawrence Erlbaum.